Our rankings of historical significance will be used to rationally assess the reputation of figures in the historical canon. But how can algorithms do this in a fair and sensible way? We start by reviewing how people approach ranking problems, to better understand the strengths and limitations of our computational methods.

Traditional Ranking Methodologies

The rankings most prominent in popular culture are produced in several ways. Perhaps most popular is the expert poll, used to create the Associated Press Top 25 College Football Rankings. Every week a set of experts (here coaches and sportswriters) independently rank the top teams. These ratings are combined using a point system, and the cream skimmed off after sorting teams by points yields the Top 25. Expert polls prove effective in clearly defined domains followed by many knowledgeable individuals. They tend to be fairly conservative, however, and are often based on second-hand judgments. Do we really believe that active college football coaches have enough time, or even a good enough cable television plan, to watch all their peers play?

Still, polls of professional historians might be used to rank historical figures. Indeed, historian polls ranking the greatness of U.S. presidents are conducted regularly as a sort of academic parlor game; an early example was the 1948 poll by Arthur M. Schlesinger (1888–1965) [27072] of Harvard University. These polls reflect interesting changes in historical reputation over time.